An ultrasonic visual sensor using a neural network and its application for automatic object recognition
Shinji Watanabe, M. Yoneyama · 2002
An ultrasonic visual sensor using a neural network is proposed and improved by reducing both the size of the neural network and the number of teaching samples. A 3-D image calculated by acoustic imaging is transformed into position and rotation invariant values, and then reorganized by a multilayered neural network. Many categories of metal or glass objects can easily be classified with this system, even when they are placed at unknown positions or rotation angles.>